Transcript: Scaling the Last Mile: Coco Robotics, OpenAI, and the Future of Autonomous Delivery

Executive Summary

In this episode of The Road to Autonomy, Zach Rash, CEO of Coco Robotics, joins the show to discuss the current state of the robotic delivery market, which he describes as being at a pivotal moment. Driven by advancements in AI, economic pressures on consumers, and labor headwinds, the demand for autonomous solutions is surging.

Zach details Coco Robotics’ multifaceted business model, which combines delivery services with a rapidly growing advertising revenue stream, turning the robots into interactive “rolling billboards.” He also offers a deep dive into the company’s technology, including its human-in-the-loop operational strategy, its key partnership with OpenAI to develop next-generation models, and its plans to scale from 1,000 to 10,000 robots.

Key Topics & Timestamps

[00:00] The Current State of the Robotic Delivery Market 

The robotic delivery market is at a pivotal moment, driven by a convergence of factors. AI technology is now sufficiently advanced to make a real-world impact, consumer-facing companies are highly focused on affordability, and significant labor headwinds such as the reclassification of gig workers—are driving up the cost of human-powered delivery. This has created both a strong need and a viable technological solution for automated delivery, making it a top priority for many companies.

[02:26] How Tipping Fatigue and Hidden Fees Accelerate Robot Adoption 

Robotic delivery offers a significant cost benefit by eliminating the need for tips, a feature actively marketed to consumers on delivery platforms. Traditional delivery costs accumulate rapidly, with menu price inflation, taxes, service fees, delivery fees, and tips all stacking on top of each other. Since tips are a critical part of a human driver’s income, and orders may not get picked up without one, autonomy’s ability to remove this cost makes the service much more affordable and accessible to a wider audience.

[04:45] The Growing Business of Advertising on Robots 

Coco Robotics has developed a growing advertising business by turning its robots into “rolling billboards”. The company runs campaigns for major brands like Netflix and Sony by applying vinyl wraps to the vehicles, which generate a high number of impressions by operating nearly 24/7 in dense urban areas. The novelty of the robots also generates organic social media engagement. Coco also offers “stunts,” where a fleet of robots can swarm an event like a movie premiere for a viral marketing moment.

[09:05] Designing the Perfect Vehicle for Urban Goods Delivery 

Coco Robotics’ design goal is to build the “perfect vehicle for urban goods delivery,” which they envision as an “autonomous bike courier” rather than a small car. The form factor is optimized to safely use sidewalks, bike lanes, and road shoulders at speeds of 10-15 mph. Its small size is a key advantage for merchants, allowing it to get right up to a restaurant’s door and enabling up to 20 robots to fit in a single parking space to handle high order volumes. The vehicle is large enough to carry 99% of typical orders, including multiple pizzas or several grocery bags.

[14:02] DoorDash Partnership and Launching in Helsinki’s Extreme Weather 

Through its partnership with DoorDash and its European brand Wolt, Coco Robotics operates in Helsinki, Finland. The city was an ideal initial European market due to its excellent pedestrian and cycling infrastructure and the high demand for delivery during its harsh winters. When severe weather makes it difficult to find human couriers, Coco can provide a guaranteed, fixed amount of delivery supply, which is a major value proposition for its partners. Operating in these conditions also forced the company to develop robust hardware that can withstand snow, ice, and extreme cold.

[18:22] Customizing the Robot’s Interior to Preserve Food Quality 

Coco Robotics works with merchants to maintain food quality by offering customizable interiors in its robots. The standard configuration includes an insulated liner and a cup holder, but merchants can also use separate insulated bags to transport hot and cold items in the same delivery. An internal camera allows Coco to provide quality control feedback to restaurant staff on how they pack orders. Enterprise brands can fully customize the experience with their own branding and specialized compartments, such as dedicated sections for french fries and milkshakes.

[21:40] Working Directly with Merchants vs. Platform Partnerships 

In addition to its partnerships with major platforms like DoorDash and Uber Eats, Coco Robotics works directly with local merchants. For high-volume businesses, Coco offers a model where the merchant can lease a dedicated fleet of robots. In this arrangement, the merchant controls the entire experience, customizing the robot’s branding, appearance, and sounds to act as a mobile extension of their brand.

[24:14] Coco’s Operational Strategy: Remote Pilots and the Path to Autonomy 

Coco Robotics operational strategy is built around a “human-in-the-loop” model using remote “pilots” who can take control of the vehicles when needed. The company started by being entirely teleoperated, which allowed it to scale and gather immense amounts of real-world data showing how human operators navigate complex urban environments. This data is used to train the autonomy system and create real-time maps of sidewalk conditions, such as construction or other blockages.

[31:44] Partnering with OpenAI to Train Next-Gen Self-Driving Models 

Coco Robotics has a partnership with OpenAI to develop next-generation self-driving models. The collaboration aims to leverage OpenAI’s advanced multimodal models to solve the most difficult driving scenarios encountered by the robots. Coco provides a large, diverse dataset from its operations in varied climates like Helsinki and Miami, including annotations from human pilots navigating situations where the current autonomy system fails. The research explores combining on-device models with larger, cloud-based AI to improve performance.

[38:08] Manufacturing and Scaling the Fleet to 10,000 Robots 

Coco Robotics currently has a fleet of over 1,000 “Coco one” vehicles, which were developed in partnership with Segway Ninebot. The company is manufacturing its next-generation robot and plans to produce approximately 10,000 units next year to meet overwhelming demand. The new model is designed with fewer parts and a more integrated supply chain to simplify maintenance and repairs as the company expands globally.

[41:41] The Logistics of Launching and Operating in a New City 

When launching in a new market, Coco Robotics partners with “anchor merchants” like grocery stores or ghost kitchens that act as initial hubs. These partners store, charge, and perform basic maintenance on the robots. Deliveries begin for the anchor merchant first, and then the service expands to other businesses in the surrounding area. To support operations, Coco establishes neighborhood-level “mini-depots” often using storage pods in rented parking spaces—so that vehicles do not have to travel long distances for service.

[45:22] The Future of Urban Delivery: Groceries, Retail, and Pharmacy 

Grocery delivery is a rapidly growing category for Coco, as its robots are perfectly sized for the smaller, more frequent grocery trips that consumers would make if delivery were more affordable. Looking ahead, once the delivery network is established, the company sees significant opportunities to expand beyond food and groceries to transport goods for retail and pharmacy at a very low cost.

[48:00] The Future of Coco Robotics 

The company’s primary focus is to capture the enormous market for urban goods delivery, which is estimated to be worth tens of billions of dollars annually in courier payments in the U.S. and Europe alone. Coco’s long-term goal is to use its technology to dramatically lower the cost of delivery, making it more affordable and accessible for everyone. Achieving this will require deploying a vast number of vehicles across many countries.

Watch the Full Episode of The Road to Autonomy

Full Episode Transcript

Grayson Brulte: Zach, it is great to have you here on The Road to Autonomy. I mean, we’ve been following the robotic delivery market quite a bit. Your company, Coco Robotics, you’re a big part of that market. In your opinion, what is the current state of the robotic delivery market? 

Zach Rash: Yeah. Thanks. Thanks for having me. I’ve been watching the show for a while. Um. Uh, it’s having a, it’s having a, a really, interesting moment right now. I think you’re getting the convergence of a couple things, which you guys cover extensively. It’s the, the technology, like the AI technology is clearly ready and starting to make an impact in the world. You first saw that with chatt and, and you know, all the LLM race, but then that’s now translated into the physical world quite a bit. I mean, Waymo went from a couple, you know, test vehicles to now it seems like they’re everywhere and you can actually ride in it, right? So I think for a lot of people, you know, um, customers, restaurants, merchants, um, and then our, you know, our platform partners, I think autonomy became very real all of a sudden. Um, and I think that’s happened at a point where. These companies are being very thoughtful of, um. Consumer, uh, consumer spending and affordability. Um, and then on top of that, you have a lot of kind of interesting labor headwinds right now. Um, like a lot of Europe’s reclassify in drivers, uh, to go from, you know, gig workers to employees. We obviously have something similar to that in California, New York with Prop 22 and some of the minimum pay guarantee. So you have a lot of pressure on price that’s gonna drive price up at a time where everyone’s a lot, a little bit more sensitive to cost. So I . both a need for some sort of automated solution. Um, and clearly the technology is ready. So I think this has just jumped the top priority list for a lot of these companies. And, um, you know, it’s hard to predict when these things will happen, but you know, we’re, we’re here and ready. And so it’s been a really cool, been a really cool year as this has become a lot more real to people.

Grayson Brulte: The technology has become very real despite. The reports, I don’t buy it ’cause I look at real world data. The consumer is hurting, the consumer is cutting back on spending. If you look at New York Federal Reserve data, the amount of debt they’re holding on credit cards is at an all time high because the consumers are, are clearly scared and so they are turning to more cost-effective. So solutions, and that’s what autonomy is enabling, especially in the delivery market. The other aspect that I believe is driving the growth of the market here is tipping fatigue. Everywhere you go, you go and buy a coffee. Oh, I want a tip. What do you mean? What did you do? I picked up at the counter is tipping fatigue, starting to help accelerate your market because a robot, you don’t tip.

Zach Rash: I, I think it’s a, it’s a good benefit to offer customers, right? When you, when you order on, you know, Uber Eats or DoorDash. After one of these platforms, um, you’ll get the option. Do you want in autonomous vehicle or do you want to human courier? Um, and if you opt into a autonomous vehicle, that’s one of the things that’s marketed to you is no, no tip or your tip gets refunded if you get matched with an av. I think it’s a huge, I mean, it’s a huge bonus, right? I mean, the way these fees work is they all kind of stack up. So you have, you have the, the subtotal, the actual goods. Then it’s inflated to account for the commissions that are being charged to the restaurant, right? So. If they have a $10 meal item, they mark it up to $13. Then you get sales tax, then you get the delivery fees, then you get the service fees, then you get the tip. The tips often calculated on that total, so. You’re paying 15, 20% on top of all those other fees. So it, it gets super expensive and obviously you want to tip your driver, especially, uh, you know, if, uh, it’s bad weather outside or whatever the conditions might be . and you often won’t get your delivery if you don’t tip the driver . for, for obvious reasons. Theri, it’s a huge part of the earnings for the driver, right? The driver needs that tip to make it worth their time. So. It’s a tough problem. Uh, you know, I think, so that’s a huge promise of autonomy is we can, obviously make this much more affordable for people. And if you do that, people order much more frequently. I think more, you know, a broader set of people can actually benefit from the convenience and delivery. I mean, clearly delivery is incredibly convenient and people want to do it, even if it’s incredibly expensive. Um, so I think making that more affordable is gonna be really good.

Grayson Brulte: Because if you look at it’s fees on top of fees, on top of fees, so you order, say. You’re based in la you order a $12 burrito. Next thing you know it’s 22 bucks and you’re like, what the heck? How did this happen? This is not worth it. That has to change. These companies have to, to streamline their fees. Autonomy is a way that can help them potentially streamline their fees. But let’s not forget, once you get addicted to fees, you don’t go back. It’s like a junkie on heroin. Once they get addicted, there. They’re not going back, unfortunately. So that’s, that’s the fee aspect of your business. The another aspect of your business, which I’d love to get your take on and how fast it’s growing is advertising, ’cause your vehicles offer all practical purposes, rolling billboards. How big is your advertising business today and how big do you see it eventually growing? 

Zach Rash: we started the ads business about a year ago . we had a lot of inbound for it. You know, people ask about it all the time. We just. Didn’t really focus on it. And then we did a couple campaigns. We did the first one with Fox, for the Grinsburg TV show. We put kind of a square wrap, uh, you know, vinyl ad on the, on the side of the vehicles that worked pretty well. Collected some data. We, we both integrated the GPS to third party trackers and, you know, did pedestrian counts from the cameras so we could give a pretty accurate, um, kind of engagement metric or, you know, justify a CPM to people we didn’t know. We just kind of made up a price and figured it out along the way . turns out we get a ton of impressions from these vehicles. Uh, uh, they’re in the busiest parts of the city, um, and they’re out pretty much 24 hours a day. So, you do get a ton of impressions. So we’ve, we’ve iterated on the model quite a bit. We did a campaign with Sony recently. We just finished one with Netflix, for the Electric State movie. That one was awesome. Um, if you saw that we had these, you know, yellow full wrapped, uh, electric state, which is about robots. So it was, it was kind of perfect. But that’s been. It’s been a great business model. I mean, I think it’s kind of like, um. You know, it’s very interactive, engaging outta home, right? Uh, you have billboards, you have bus wraps, uh, but robots are novel. Everyone films them, uh, you know, they always end up on TikTok and Instagram posts. You get a lot of indirect impressions that way. But it’s a really cool experience and it’s cool way to bring the, you know, the ad to life for people . offer this thing we call stunts, which is where, you know, you could have a premier in Hollywood and we will pull all the Hollywood vehicles for, you know, 10 minutes and they’ll come swarm the event and it becomes this kind of viral moment, and then they’ll go back to doing deliveries . that kind of stuff’s really cool because you can create a very experiential type of marketing campaign around it. Wherever people are having events, usually we have tens of vehicles anyways, right? I think we have probably 40 or 50 vehicles in Hollywood, so they could just. Swarm around, uh, the premier, which is pretty cool. So it’s a growing part of our business. It’s, it’s, um, yeah, obviously great, great margins for a, for a delivery company, um, as all the delivery companies have figured out. Advertising is a pretty important component . but I do think it’s going to become. I think it’s actually a really valuable form of advertising, um, especially as you start getting a lot of, uh, disruption in the digital channels right now, um, and how this all unfolds with AI and the chat bot interface. Um, I think this is a really, is gonna be a really strong and really durable form of advertising.

Grayson Brulte: It is just this tangible, the consumer can see it, and you have those, those viral moments that get picked up quite a bit and your advertisers aren’t paying extra for that, so that’s a benefit for them. Is the largest advertising base. Is that the The film studios? The media companies. Is that your largest advertising base today? 

Zach Rash: We, we, we’ve worked with others and we’ll continue to work with others. Uh, it was the most natural, organic thing for us to do, being around all the studios in Los Angeles. And, you know, they’re, they’re promoting a show or a movie, so it’s usually like a big event that they want to have kind of appear everywhere, um, before the premiere. So, that was just the natural place to start. But what, you know, we’re, we’ll, uh, we’re certainly working on, um, more different, uh, more categories.

Grayson Brulte: Do you see a path where your advertising revenue becomes larger than your delivery revenue? 

Zach Rash: don’t know. We’ll, we’ll see. I think kind of what I was saying about the AI chat bot world, it’ll be really interesting to see where this goes. Because I mean, those ad dollars still have to get placed, right? So if ChatGPT is not gonna monetize them for search or it’s not gonna monetize as well or is the same way, and you know, these AI chatbots have the. Balancing act of they want to represent truthfulness, in a way that, you know, a Google list of blue links doesn’t exactly have to do, it should be relevance. But if, if the AI is supposed to represent truthfulness, it’s tough to incorporate an A into that flow, um, directly. I mean, there’s a bunch of ways they’ll do it, and I’m sure they’ll figure that out, but. Those same ad dollars have to get placed somewhere. And so the, the, the CPMs will just go way up on most of these traditional digital channels, if that happens. Um, so then outta home might be a very, very durable way to do this . so I, we’ll, we’ll see. Um, I think there’s a really cool world where we monetize the, you know, the, the advertising and the, the size of the vehicles. And that off allows us to offer a much cheaper service than we would. Be able to do previously. Um, so I think that could be really interesting and that could really benefit consumers and merchants and that can grow volumes. And I think all of that is like net beneficial to the company. So we’ll see how it shapes out.

Grayson Brulte: And as you grow, do you see different potential form factors for your delivery? Robots emerging in the future.

Zach Rash: So what, what we, what we’re designing at at Coco is we’re trying to build the perfect vehicle for urban. Goods delivery. Um, so you have the kind of a Waymo or a car type vehicle, um, that is not the best form factor to move things around a city, right? There’s a reason that most cities deliver things with bicycle couriers. So you could think of us as building an autonomous bike courier . that’s, that’s, you know, in the short to medium term, that is our, that is our North star on the hardware. Um, how do we use bike lanes, shoulders of roads, sidewalks? How do we kind of get to that 10, 15 miles an hour, but still be able to fit safely and appropriately on the city sidewalks? that is the kind of ultimate design goal, and that’s really important for our merchants because you want to be able to get right up to the door of the restaurant or the side window or the back alley or wherever they, um, wherever they want to put the food into the robot. We can fit 20 of these in a single parking space. So if you have a merchant that does a hundred orders an hour at peak, that’s really tough to do in a city if you have a larger vehicle . so, and the vehicle’s big enough to fit, you know, 99% of orders you could fit like 40 bowls, uh, you know, like a sweet green bowl or a, you know, some sort of bullet in it. Uh, you can fit 10 extra large pizzas. We can fit four Grocery bags . it is large enough capacity wise to deliver most of the things people get delivered. It’s super convenient for the merchants because it’s small and can fit on. The sidewalks can handle high volume . high throughput, and then it can get right up close to the customer. The customer doesn’t have to come out and try to walk a couple of minutes to find wherever the vehicle was able to pull over. But you still wanna be able to get off the sidewalk to get higher speeds so you can deliver further and you can deliver faster. Right? Going faster is always gonna be a good thing. Um, but you can imagine a place like New York or downtown Chicago or you know, um, Miami these days has a ton of traffic. Um, a car is not actually the best way to get around that city. Um, in particular if you’re delivering goods.

Grayson Brulte: There’s some scenarios where a New Yorker. I don’t take a car. I walk everywhere downtown Miami and Brickle. I’ve gotten out of Ubers and just walked. ’cause it just takes so long. There’s just, there’s so much congestion.

Zach Rash: it’s gotten pretty bad.

Grayson Brulte: It’s horrible. I don’t like traffic. I like to walk. I i’ll, I, but in the summer it’s hotter than hot. But I still rather walk and sitting in the car getting frustrated and as somebody says, doom scrolling. ’cause that’s what you do when you’re bored in a car. It’s annoying. Hypothetically, you’re at a, at a merchant, let’s just use a restaurant for an example, and there’s 20 of your Coco bots lined up in one parking spot. How does the individual loading those bots know which one to go to, to put what delivery into what bot? 

Zach Rash: Yeah, we have um, we have a first order ready, first order outflow for merchants that do that kind of volume. So you can think of it, it kind of becomes like a taxi line. Um. So they can actually assign any robot to the order . so, um, we actually have a, we have this example. There’s this merchant called Main Chick Hot Chicken. Um, they operate out of a ghost kitchen in West la . they prep, they’re super consistent on their prep times, so you know, the amount of time it takes ’em to get the food ready and bagged . and that’s really quick. Um, and that’s the only thing they make is these, you know, spicy chicken sandwiches. And they’re, they’re really good and they’re really consistent at making them. So we didn’t experiment with them. We said, okay, if any robot, if any order is eligible for robot delivery, why don’t you prep it first? So you basically, uh, pop it to the top of the queue . and because, you know, we are ready and available, they do so much volume. They have a whole fleet of vehicles sitting there. And that actually was able to cut down the delivery time for the customer dramatically. Because the second to order’s Ready, it went into a robot and the robot was off to the customer. I dunno, that’s 15, 20 minutes. Assigning a driver, having the driver park, having the driver come over and look for their order. They might be picking up two orders. They might be using two apps at the same time . and so, uh, this can save, you know, I. 15 minutes there, right? So obviously if it’s a three mile delivery, we’re gonna start losing that advantage because we’re slower than a car, and Los Angeles is a lot of car based delivery. But if the delivery’s going a mile, we’re gonna, we’re gonna be way faster, a mile and a half. We’re gonna be way faster than a, than a, you know, a human in a car. So, that’s something really cool. We’ve worked on with these higher volume merchants that flow doesn’t make sense for 99% of restaurants, but if you do hundreds of orders a day, um, that’s been the best flow, um, both for the merchant experience and for the customer experience.

Grayson Brulte: What is your average delivery length? Is it under, under three miles, mile and a half? What does that look like for your average distance? 

Zach Rash: Yeah, about a mile and a half we do, we deliver up to two miles, um, travel distance . so the average delivery is a little over a mile. Um, it depends. All the cities are different, right? You have, like Miami and Chicago, we’re a little bit denser than Los Angeles. Helsinki’s, dense. Um. Helsinki’s not even, you know, helsinki’s like low density for a European city, but still it’s Europe and Europe’s just built way denser than the most of the United States . um, you know, almost every single delivery in Helsinki’s, uh, within a few miles.

Grayson Brulte: In Helsinki that you did that through a partnership with DoorDash, and as part of that partnership, you successfully completed a hundred thousand deliveries. That’s really impressive. How did you do it and how did the partnership with DoorDash come together? ’cause eventually that partnership you exported from Helsinki to the United States where you’re now going to Chicago with DoorDash.

Zach Rash: Yeah, so we’re, we’re in Chicago, Miami, um, Los Angeles and, and Helsinki with DoorDash. Um, yeah, they, they’ve, they’ve been, they’ve been a great partner. We, you know, partnered with both them and Vault, which is their, their kind of European, um, brand. Now, uh, they’re based in Helsinki, so that’s why we started in Helsinki. Helsinki was their first European market, um, because that’s where the, the team, the Vault teams headquartered . and I, you know, at first I, I, I wasn’t really expecting to go to Helsinki as the first European launch. You know, everyone always asks me, you know, why, why did you start in Helsinki? Um, Helsinki’s actually really great in a lot of ways. I didn’t expect the infrastructure’s perfect. The bike lanes are on the sidewalks. The sidewalks are like 10 feet wide. So traversing the city as a pedestrian or a cyclist is 10 outta 10. And that’s one thing we care a lot about. The weather in Helsinki also makes it so that in the winter everyone wants to get everything delivered it’s, you know, it’s dark, it’s dark all day, it’s cold, it’s, you know, it’s like negative 15 degrees and it’s, uh, snow everywhere and ice. Like nobody wants to go outside. Even if, even if all the restaurants are like three blocks away, nobody wants to leave their house. Um, so, but at the same time, couriers do not want to be on the roads. It’s mostly bike couriers. It’s a very hard time to get drivers on the road, obviously, because it’s miserable outside. So. Um, you kind of have this, this problem where when customers want it the most is also the times where, where it’s hardest to get supply on the road. So we can offer a lot of value as an autonomous vehicle company to say, you know, we can give you this fixed amount of supply. You can plan ahead and say, I want, you know, I want 10,000 supply hours every day, guaranteed from eight, you know, 7:00 AM to 3:00 AM . and we can guarantee that no matter what the weather conditions are, so the, the poor weather to Helsinki actually is a huge selling point to our customers. And then it forced us coming from Los Angeles to really develop the hardware, to be able to handle super low temperatures, ice, snow, and that whole kind of workflow and those conditions.

Grayson Brulte: You go from, it hits a hundred in LA and if you, Johnny Drama, when I remember he was going over the hill in an episode of entre just going higher, it’s going higher and it gets, can get 120 in the valley. So that’s one extreme temperature. Then Helsinki, you get minus 15 Fahrenheit, minus 2020 Fahrenheit. How do you develop the packaging? So in, let’s say in Helsinki, for example. I want my pizza warm. Where in LA you’re delivering me ice cream and I want my ice cream cold. I don’t want to have melted cream. How do you build the packaging for these different environments? 

Zach Rash: we work with the brands, you know, we have a kind of a generic robot layout where you get this, um, insulated, uh, interior liner in the robot. There’s also a cup holder attachment, so that can fold up or it can be folded down so you can put drinks in it . the standard you get, you know, our average delivery times travel times 15 minutes. So it’s, it’s not in there for that long. That works pretty well, even in the, even in the, um, you know, harshest weather conditions. We also offer merchants is you can separate hot and cold. So if you say, I wanna deliver this pizza and I want this bottle of soda, or I’m delivering this pizza in a salad, the pizza will steam up the whole compartment and it will keep it very hot . so you don’t want that, it will heat up the bottle of soda or the salad, so you can actually put that in a separate insulated bag. So we work with merchants to customize this. This is a huge selling point of using robots for delivery. Your guest will receive their order exactly as your staff puts it in. And we have a camera on the inside so we can also help you with quality control to understand are you putting it in there correctly, right? Are you, are you stacking the heavy part of the ramen on the top and it spills all the time, right? If you’re doing that with us, the same thing is gonna happen with a driver. Um, so we can actually do quality control on the best way to deliver that food. You can customize that experience too. We can work, we work with enterprise brands to really own that, so you can put your logo, your branding on it. You can say, I want a french fries compartment and a milkshake compartment . and so you can really go above and beyond. Um. that guest experience that’s customized for you. Um, and I think that’s really cool and that’s something that’s been totally lost in delivery. You know, we used to do that when we had the local pizza delivery guy, right? He, you know, it was like a staple of the neighborhood he lived in the neighborhood. You’d invite him in and the kids would run down and say like, oh, the pizza man’s here. Like that doesn’t happen anymore. And that was a huge part of the pizza delivery service was that person was supposed to be. Um, part of the experience, and now everything’s just kind of left at your doorstep. ’cause, you know, we’ve, we’ve scaled that industry a lot. So I think this is a cool way to bring back that kind of human and that hospitality, um, ironically through, through a robot. And I think that would be really good for, for, you know, creating delightful experiences for The customers.

Grayson Brulte: The pizza man was, was part of the culture. Every, everybody knew him. And then remember back in the day, the Domino’s 20 minutes or, or less. Guaranteed or it’s free. And, and you hear these stories about all the crashes and all the different scenarios that happened there, that that’s pizza delivery. Pizza delivery’s evolved. They, they now, they leave it and they’re not here 20 minutes or less. Thank you trial lawyers for ruining that experiment. But Domino’s gained a ton of market share from Pizza Hut, and that’s how it started. To give you a little background on that, pizza Hut didn’t offer delivery, but they guaranteed it within 15 minutes to pick it up. Domino said, well, we’ll deliver it within 20. Next thing you know, pizza Hut goes into bankruptcy. And domino scales a very, very large business. Can your robots pass the French fry test when I order from a restaurant? French fries. Now you get ’em. They’re sake, they’re horrible. Can your robots deliver the french fries when I bite into it? Ah, this is crispy. Can they do that? 

Zach Rash: The, so the way to think about it is our vehicle is gonna be as good as possible at delivering the food based on how the merchant uses it. So if the merchant, if the merchant the french fries in, in a bag, or they have ’em in a container inside the bag that doesn’t let steam out, they’re gonna be soggy . the, the, the benefit of the robot is we can work with you to say, what is the perfect packaging? What is the perfect way to do that? And that way we’re kind of a consultant to the merchant, but at the end of the day, they are responsible for using the vehicle to their, to their benefit, right? We get merchants that don’t use the cup holders and they just put the drinks in there. It’s like, you know, that’s what we’ll give you that feedback, but, you know, you should use the cup holder. Um, or you’re gonna, you know, mix up the coffee and stuff. So, You know, so, so we, we, we advise and we have the camera data so we can actually give them feedback, which usually there’s a kind of a mismatch between what the owner wants to do or the general manager wants to do and what might be happening with the staff. You know, the restaurant industry is tough, they’re busy, get high turnover . so we provide that data to the restaurant managers so they can kind of manage quality, but we’re only gonna be as good as the restaurant. Um. You know, wants us to be, uh, but we are very involved in that and obviously like more enterprise customers take this stuff super seriously, but trying to give that same technology to the SMBs.

Grayson Brulte: Yeah, I mean, to, to a restaurant, the restaurant’s a brand. You have to have a consistent experience, whether it it’s in, in, in-house dining or out of house dining. You have to have a co experience where your business over time will collapse because the restaurant business is a very. Finicky business. You had the partnership with DoorDash. You also have one with Uber East. Outside of those two partnerships, if the local restaurant, let say the local burrito place that I want to order burrito and I want to deliver it autonomously, how do I do that? 

Zach Rash: yeah, we work with, Uber Eats. We work with DoorDash and Volt . work directly with, um, yeah, merchant, merchant partners, so they can reach out to us through our website or they call us, contact us in a number of ways. Then we’ll assess if it makes sense, right? They might want one vehicle that offers some sort of baseline capacity, and then they can kind of flex up and use, uh, you know, drivers, um, to kind of meet a higher volume, uh, of orders . you have enterprise restaurants that might only deliver a mile or two, and they want only robots for all of the orders. . So both, both of those are fine. We’ll work with them on what makes the most sense. The economics of those things are different, right? If you’re using a robot that’s just gonna be busy 20 hours a day, that’s very different than, you know, having enough robots to meet 7:00 PM peak. , But fundamentally the cost structure’s way lower than, than having a staffed human driver, right? So if you want to offer that experience and you only want robot, the robot delivery experience, um, you can do that. Um. So kind of calibrate you on the price point based on our cost. So we have, we have a couple partners doing that model where it’s robot only . and um, yeah, it’s a little bit different of a business, but it, but you know, that our cost structure low enough . it can actually it can actually work.

Grayson Brulte: You sell or lease those robots to those dedicated restaurants.

Zach Rash: Yeah. It’s, it’s, it’s kind of like a lease.

Grayson Brulte: Wow. And so that’s their, that’s their fleet. Can they put their own advertising on it? Or they have to take your advertising? 

Zach Rash: They can, they, they own the branding, they own the experience. We will build, we will build that fleet to do exactly what they want. That can, we can customize the interior for them, the exterior, the sounds, you can kind of sound, you can brand it however you want. That should be an extension of your, of your store.

Grayson Brulte: Is that a growing business for you inside of

Zach Rash: Yeah, that’s, I think that’s gonna be a. Large focus for us. Um, it doesn’t make a ton of sense for most But you do have high volume brands, and that makes a lot of sense, right? If, if you, if you wanna work directly with us and you’re a small business that does one order a day or something like that, you can pull from the fleet. Like, we’ll still fulfill that order for you and it’ll be a per delivery charge. Um, we’ll still work with you on your own channel, right? If you want to in phone orders and then fulfill it with a robot, you can pull from any of the other, you can pull from the fleet that’s roaming around the city and deliver with that. If you want to like really own and customize that experience, that’s more of like an enterprise type model. Um, then it’s more of a fleet lease and you’re responsible for, um, you know, utilization of that, of, uh, of the vehicle. And we’ll work through the kind of capacity planning with you and all of that. But yeah, I mean, I think that’s the, I think that’s gonna be a huge business for the, um, for the future for these brands that really want to kind of take control of that experience.

Grayson Brulte: From an operational standpoint, how does it work? Because you’re operating a hybrid fleet where you’re first remote driver, and then autonomy as a, as a fallback. So how does that operate from an operational standpoint? 

Zach Rash: Yeah, so I think, you know, every autonomous vehicle company has somewhere, has some sort of tele operator, some sort of remote operator in the mix. What we did at Coco from the very beginning is we treated the. You know, tele operator job, we call them pilots. We treated the job of a pilot as kind of a first class product, um, inside of, inside of Coco. So, we’re very thoughtful about all the software on how do you dispatch the right pilot to the right order? How do I switch pilots between different countries or different cities? How do I very quickly present them all of the context? Where am I? What are the rules of the road? What should I watch out for? Uh, you know, what state of the delivery am I on? Um, what, what is needed, um, for me? Like why did the autopilot disengage? Um, so you have to be really good at building all that. And we started saying we’re gonna be entirely teleoperated. Um, so that product had to work really well because that did 100% of the deliveries. Then over time the vehicles have been trained to be, uh, you know, a higher degree of autonomy. And then when we do have to involve an operator because we run into some sort of construction zone or some sort of anomaly, or we’re crossing a really busy intersection and we wanna have like the redundancy of a pilot supervising it, that product’s already built and works really well. And the other thing I’ll say is like, if you want to train this autonomy system, you need a lot of data. And so our intuition was always like, well, let’s make, let’s build inexpensive hardware . a really good product for teleoperation and let’s start scaling it up. And now we collect a ton of data in the most chaotic urban environments, you know, possible, right? We’re operating 20 hours a day in the craziest parts of the cities in some of our busiest cities around the world . get a ton of data, and it’s not just the video data, it’s the what did the human driver do? How did they interact with this? Um, they also annotate our maps, right? So we have a real time map of every sidewalk and every city we operate in. So I could tell you where every car is parked blocked in the sidewalk, where every kind of homeless encampment is, where every um, uh, uh, yeah, construction zone, uh, streetlight, that’s out. Um. You know, we, we record, you know, a bunch of these different events using that human operator, that also is very useful training data because you do need to know a real time map of infrastructure to be able to navigate the city, the city well, right? We’re not primarily using roads. And so that same infrastructure that exists for road conditions and real time updates on that doesn’t really exist on the sidewalk. So we had to create that . but once you have. That sort of, um, data collection engine built into the company, autonomy becomes much easier to roll out because you have all this as training data and you have the ability to pull back in the driver to complete any of these data points, narrate what’s happening, um, confirm that it is indeed a blockage, et cetera, right? You’re always gonna wanna use them. You just wanna be very efficient at how you use the, operator.

Grayson Brulte: What is the ratio of human operator to robots in the field? 

Zach Rash: So the, the, the way we look at it is you, you basically want one operator to be supervising, you know, multiple at a time. And so it varies depending on the city, depending on how, how mature that market is, how much data we have on that market. Right? Helsinki looks very different than Los Angeles and Miami. We’ve started collecting snow data just this last winter was the first time we actually had serious snow data, right. Chicago and, and Helsinki. So, depends on the market, but the idea is you have, at some points you have, one driver driving a robot a hundred percent of the time, um, in certain conditions. And other times it might be supervising multiple at a time. , And, um, I think this is also an under, maybe an underappreciated part of, uh, of why this form factor vehicle . become so low cost quickly . if you have an operator watching three robots at a time, for example, that’s still fully supervised autonomy, but that’s a huge efficiency gain for us, right? That’s three orders happening simultaneously with one person. , And then you go to four, you know, four robots at a time, five robots at a time. You know, at some point you can’t supervise more than, you know, four or five at a time, or there’s just too many feeds to watch. If you’re traveling at 5, 6, 7, 8 miles an hour and you’re watching, you know, three feeds at a time, it’s still supervised. So you can still be on top of catching kind of false, uh, false negatives where it might miss something that it was supposed to stop for. So you have a very high engagement rate with the, with the pilot, but you’re still getting a lot of the economic benefit that helps train the system even faster, and then you can improve that ratio and then you have really good data to say. We’re happy with our false negative and our false positive rate to the point where we are happy to staff one to 10. Um, and then, you know, in these domains we can be a little bit more lights out. So that’s what we’re working towards. That’ll be a kind of a continuous evolution for the whole company’s history, I’m sure. ’cause you’re never gonna be fully lights out. No humans. It doesn’t make any sense. Um. But the goal is just to be really efficient without impacting delivery times and while being, you know, having a, having a, a very safe operation. Um, the last thing I’ll add there is, this is also I think, underappreciated . you’re trying to deliver two miles and you’re traveling, you know. 10 miles an hour or less. You are very, very sensitive to any downtime, right? You do not wanna stop. Um, and, uh, you do not wanna have to reroute. So it’s not just that, is the vehicle capable of getting from A to B autonomously? It is, is it capable of getting from A to B as fast as possible and your tolerance for any hit to speed or time spent stopped waiting for an operator to come in and fix something or making a mistake? That cost is super high . you want to be very thoughtful about rolling out autonomy in a way that has. Does not impact your delivery times. Um, and in this world, it’s not really your average delivery times, it’s really about your 90th percentile or your 99th percentile delivery times. So you wanna say, okay, in the worst kind of conditions and the hardest delivery we do that day, you know, how, how delayed was it because of autonomy? And, you know, the answer can’t be 20 minutes. Um, that’s a canceled order. That’s a terrible experience. Um, we’ve failed that and we shouldn’t have released it.

Grayson Brulte: It’s the goal to get to full autonomy where you say have an intervention every a hundred thousand miles, every 500,000 miles, and that way you can truly scale the business globally.

Zach Rash: For sure. I mean, the goal is always gonna be to have as few interventions as possible, and to have the system run at a very high level of autonomy. , So that, that, I think that’s gonna be the goal for, for any of these companies. I. I think you know, the trying to make is you get a ton of economic benefit along the way and that end state. , Is not gonna be too economically different than having like a, a one to 20 or one to 30 or one to 50 ratio where you have one person who’s kind of supervising, uh, or responsible for 50 different, , vehicles at a time, or 50 vehicles where they’re kind of coming in and completing tasks. I mean, that’s getting pretty low cost. , And that has, that human interaction has a ton of value to our system. So maybe eventually it gets good enough where you don’t have to do that. I think you’re always gonna want a human in the loop for some things. , So I, I think it’ll be a continuous, a continuous journey to get there, but that human feedback is incredibly important.

Grayson Brulte: It helps your system learn and then over time as your autonomy stack gets us called more and more mature. Do you ever see a new business emerge where perhaps you license that technology, your competitor Serve robotics is now licensing it to Magna? Do you get to a point where perhaps you license your stack as well to say a, a Continental or a Bosch? 

Zach Rash: the closest thing we have to, that . not licensing anything per se, but we, we do have partnership with OpenAI. So, we’re working with OpenAI to develop, you know, next gen models for self-driving, right? They’re very interested, obviously, in getting their large text or image video, their multimodal models actually doing useful things in the real world. Uh, the real world has a lot more data than the internet, and they’re very interested in how do we make our models useful in the real world so that they can. Do useful things, but also improve the models generally. I think we’ve seen from these large models, um, making it better at any new task that makes it better at all tasks . and so, um, that partnership’s really cool ’cause we’re exploring, you know, how, how does, how do we get these models into the real world? Are they useful? Um, what’s required to make them useful and can we kind of work together to make them better? Um, so that’s still very early stages of that partnership, but we have, we have, we have a lot of data and a lot of human, um, annotation on that data as I pointed out, and in kind of the most diverse situations, I think a lot of AV companies have usually been California, Arizona, you know, good weather deployments. But we have very bad weather deployments. I mean, we have now Miami and Helsinki, Chicago, so you’re getting a pretty big contrast in that data . and how does a human operator navigate those? critical areas where our autonomy algorithm cannot succeed. And so can we work together by having a, you know, a model running on the vehicle and a more cloud-based, larger model? Can those work together to solve some of the hardest driving scenarios? And I think that’s a really interesting part of robotics to explore.

Grayson Brulte: It’s a very interesting part. Do you potentially see. The open AI data set merging with your data set to build an autonomous driving stack, perhaps in the future.

Zach Rash: So I, uh, one of the thoughts we had is if these models keep getting smarter and smarter as they have, you know, seeming to do every six months, they just seem way more capable and you have to play that trend out. You know, there’s a very real possibility that these models drive can drive a robot or drive a car super well. Um, they’re trained on the entirety of the internet. Um, they’re, these companies are spending tens of billions of dollars to do this. There is, uh, an arms race across all of big tech. So this is unprecedented stuff and the models are just getting better and better. And we haven’t really seen the capabilities plateau. So, you know, it’s not a certainty, but we do want to be. Very thoughtful of where we’re spending our time, where we’re spending our energy with that trend line, looking the way it does. So if, if those models become super capable, right? We, we want to know, we want to be the first to know, we wanna be the first to integrate it. We wanna be the first to use that to, to. To make the, the service better for our customers. I think we’re quite a ways away from those models being able to like fit on the device and drive them, right? We’re not gonna put, we’re not gonna put tens of thousands of dollars of compute on these vehicles. That’s not gonna happen . you’re always gonna need some combination of custom, um, custom, uh, stack that runs at a high frequency on the robot. But I think being able to use these larger models in the cloud, I think is gonna be really powerful.

Grayson Brulte: It’s gonna be very interesting because I’ve. Red papers and scientific debates around using the Meta Llama models potentially to power an autonomous driving. Obviously we see what, what alphabet and Waymo are doing. There’s probably some deep mind in that code. We saw the EMMA paper, so there is clearly a a path there from, you’re right, you’re not gonna load a bunch of Nvidia GPUs on these things and say, giddy up. Here you go, because. Um, and use the term, you’re gonna get a cowboy that’s gonna gonna pick ’em up and steal ’em and rip the GPU out of ’em because of the, the value there. And perhaps they end up on some crypto mining farm. Is the onboard commu, are you running low power arm chips on there, or, or what? What are you running for the onboard compute today? 

Zach Rash: Yeah, we’re using the Nvidia jet boards. , They’re super low power. It’s got, you know, uh, it’s got unified memory across GPU and the CPU. Um, it’s super low power, um, but it’s really capable. I mean, those boards are getting really good. Um, know, the first gen vehicle we had all had the, um, the Xavier boards and now they have the Orin boards and, um, it’s, it’s super capable. They’re, they’re low power, low cost, um, so. Plan to continue to use that, uh, that line. But, but yeah, then you wanna be able to selectively say, okay, crossing this 12 lanes of traffic in Chicago and the lights are out and it’s a storm and I can’t really see very much. Am I safe to cross now? right. A question like that, our model might not have the capacity to answer that and. A very large cloud-based model that’s been trained on all of YouTube or something might, might be, might have some patterns that are relevant there. Um, things to look for, um, a ways to do risk assessment. Um, uh, and so I think that’s, I. I would be surprised the future doesn’t go in this direction. All of our models are also trained end-to-end. So we just take all of the video data and all of the maps data, and then all of the pilot data, which is both, you know, any voice narration, they do any, any annotation they do combined with their trajectory that they drove. And we just train a model end to end to say video in output. The next, you know, 10 actions, um, So that fits really well into this paradigm. ’cause now you’re, you know, these kind of architectures and these, uh, strategies are all kind of merging, right? Because that, that also just looks like tokens and tokens out. And so, um, I think as these kind of paradigms merge the LLM paradigm with the self-driving paradigm, I think we’re well on our way to do that. Um, uh, then I think. The, the advances in one will continue to benefit the other. So I think that’s gonna be really exciting. And there’s unprecedented amount of, you know, spend and research going into this right now. So it’s an exciting time to be in robotics. It’s gonna change very quickly.

Grayson Brulte: It is at a very exciting time. How many hours would you estimate a day or a week are you putting into the model? 

Zach Rash: we retrain, We don’t typically just retrain on everything. We’ll retrain on the kind of hardest examples or where, uh, disengagement happened or, you know, we have a bunch of tools to query the fleet to say what’s interesting. So I don’t know. It’s the, uh, probably, probably in the tens of thousands if I had to guess something like that. I mean, we’re training these things on millions of hours.

Grayson Brulte: If you look at your open AI partnership, you’re training on millions of hours. That gives open AI potentially a way to counter alphabet’s vo product. ’cause they’re training that off of YouTube data. Perhaps they could train off of your, and put a, a video product out. So not, not only can you create a better self-driving car, I. They could potentially create a competitor to Alphabet, which where this all goes is it’s getting really, really interesting. We’re gonna see all types of different mergers evolve, different models of evolve, but end to end, as you clearly know, is the future of where this is going on the hardware side. So you, your model scaling on the software side, on the hardware side. Who’s manufacturing the robots today? 

Zach Rash: So the, the generation you see on the road now, we call Coco one. Um, Coco one was made, um, with a couple different cms. Um, the, the, the one we talked about is, uh. Segway nine bot . they made, we, we partnered with them back. This was back in 2021 and we partnered with them because they made like all the scooters. Um, and so the idea was, let’s take, let’s, you know, let’s capitalize on the supply chain that got created to support millions of scooters and all of the r and d that went into making the, a very reliable outdoor, you know, power electronics. Uh, so, that was the idea that we made that base with them, like the kind of chassis, uh, with them. And then we basically just made this big injection molded tub where we mounted all the components. So we have, um, we have over a thousand of those. Um, and we’re making the next gen vehicle right now. Um, we haven’t, we haven’t talked about who the, who the partner is yet, although I’m sure we will soon in the future. Um, uh. But we’re, we’re probably gonna make next year, probably gonna make about 10,000 of those . it’s gonna be similar, It’s gonna be the usual things like fewer parts, easier supply chain to manage, um, for repairs, right? If I have a fleet all over Europe, in the US you don’t wanna manage like hundreds or thousands of parts. You want kind of, uh, more vertically integrated, simplified electronics, um, you know, all the, all the kind of, uh, basic stuff. Once you get to a little bit more scale, you would just build things to be a little bit, um, uh, easier to manage on the supply chain front . and then, you know, suspension, faster. Um, uh, directionally, it’s, it’s gonna be, it’s gonna look pretty, pretty similar. We’re, we’re pretty happy with the, the current vehicle we have. We just need more of ’em. And then there’s a bunch of stuff we can improve if you’re starting from scratch.

Grayson Brulte: 2025, you have a thousand today. 2026 you’re going to 10,000. So we’re looking at a 9,000 bot scale in next year.

Zach Rash: I, we’ll see the exact ramp rate, I think 10,000 by the end of the year. So we’ll see how many we actually can get deployed and operationalized in that timeframe. But yeah, that’s, that’s like roughly what we’re looking at here. We, we have way more demand than supply right now. We’re completely supply constrained on the number of vehicles. Um, so we’re, we’re, we’re trying to, Trying to put out vehicles as fast as physically possible. Um, uh, we have a, we have like a whole, we’re doing a lot of assembly in Los Angeles right now, um, to get more out. So we have like a full tent, our whole parking lot’s fully tented up, uh, trying to, trying to push out more vehicles so it’s all hand on deck to get more supply hours out. Yeah, we’ll, we’ll be that next gen vehicle will, will arrive kind of early next year.

Grayson Brulte: How do you put ’em into the market? So you’re in Chicago, you’re, you’re in Los Angeles, you’re in Miami. Obviously you and I both know that there’s more markets coming. When you get that 10,000, how do you deploy them in the market? And then where do they go to sleep at night? Where, where do you have to set up operations and depots, new city, or how does that look like and how does it work? 

Zach Rash: So we’ve spent a ton of time making sure that the business, like the core in economics, are profitable today and profitable in the new city. So we, we. We make money on all of our deliveries. We, we, we have a, like, fundamentally profitable business model today. We’re not waiting for hardware costs to come down or anything like that. The vehicles are, are inexpensive. , And the total cost of ownership really low. And so we got that model working in Los Angeles, but you know, we all live here and we have an office here that is, it’s also a warehouse. So, um, you know, a little bit more straightforward to do so then when you open up new markets, we basically start by partnering with what we would call like anchor merchants. So these might be merchants, uh, like larger enterprise merchants. It could be like a, a ghost kitchen company. Um, it could be like a Grocery store. Um, and so we will actually ship the vehicles to them first, into a new market. Uh, they will store them, they will charge them, they will clean them, they will do basic maintenance. We’ll store. spare parts there. Um, you know, going back to coco twos to make sure that spare part, um, uh, you know, number and cost is, is as low as possible ’cause you do have to store it locally . so we’ll ship all that there, and then they get dispatched from that anchor merchant. Um, then we, you know, we’ll start mapping the area, get ready for operations, make sure we understand how to get from A to B, um, super reliable. We understand all the quirks of the city. Once that mapping process is complete, we’ll start doing those orders for that anchor merchant. And then we’ll start onboarding the merchants, um, in the surrounding areas and then expanding from there. And then, you know, as we start getting more and more merchants, you need more storage locations, more charging depots, um, uh, et cetera. So, um, but we try to do this mostly with merchant partners . if there’s an area where we don’t have a merchant partner, we can do this with. We will rent parking spaces from the city or from a business. We’ll put a storage pod there and the robots will go back there at night to charge store spare parts, right? You basically want this like mini depot, at the kind of neighborhood level so the vehicle do not have to be moving around the city.

Grayson Brulte: So if you look at, let’s use la. I’ve lived there half my life. You’re based there and LA is vast. So do we have a, a hub, say, in Brentwood by the country mart? You have another one in say, downtown Venice and perhaps you have one somewhere in the Golden Triangle in Beverly Hills. And is that how you kind of map them out? 

Zach Rash: Yeah, so they’ll, they’ll, you know, stand, there’s a Santa Monica one, there’s a West LA one. It’s like, you know, kind of near the 4 0 5. Um. We have a West Hollywood, one, Hollywood downtown, Mid-City . so all the kind of major hubs will have a place where the robots can go back to, um, sp don’t have to get transported around the city. I mean, going, you know, downtown to Venice could be like two hours at the wrong time of day. So you want to minimize any of that transportation. So you want to, you wanna do repairs, maintenance, all of that. You want to do it as close as possible to where those vehicles are deployed. You want to have them be kind of charged, uh, without, without a lot of human involvement. Um, and then you want them clean, right? Like they’re not, not, you need, you wanna store them in a way where they’re not gonna get dirty and you don’t have to go clean it every morning. You want them to, you know, be clean and presentable, um, uh, ’cause they are delivering food. So, you know, it’s important that they look good. Um, so we work with the merchant, a lot of our merchant partners to make sure that that stuff’s complete right Vehicles, clean vehicles, charged vehicles, you know, out on the curb. Um, this is, I think, an important and underappreciated part of the hardware design as well. Um, it smaller makes all of this way easier and cheaper, right? The larger the vehicle is, the more complex a lot of this stuff is to do. And then obviously if you have a full size car, then it’s, then you can kind of . the vehicle, um, drive around a larger zone and traverse the whole city. Um, but as I’ve pointed out, uh, going a full-size car I don’t think is the right form factor for urban deliveries.

Grayson Brulte: No, it’s not. And, and on that front, we, we talked about restaurants, but another category that you’re very. All with is Grocery and for Grocery stores they have parking lots and they have the trucks go in. So perhaps back where the delivery bay is, you could potentially see a pod going there. It just goes right into their operations.

Zach Rash: Yeah, we have, we have a, we have an example of that. We have, um. Gro Grocery, Grocery stores are a perfect fit for that. ’cause they usually have a lot of real estate. And yeah, gro Grocery been a very fast growing category for us. Uh, like I said, the vehicles fit about four Grocery bags, which I think is the perfect size because as you get much larger than that, the economics and Grocery delivery start becoming. It’s pretty good. Uh, right? Someone orders like $300 of Groceryeries, um, you know, charging them $15, $20 for a delivery fees, not the end of the world. But if that’s totally cost prohibitive, if you’re ordering $30 of Groceryeries or $40 of Groceryeries. And so I do think in the future people would like to shop more if, if you had the convenience delivery. I think people would like to shop more frequently, smaller baskets, more closer to when they’re actually gonna prepare the meal. Um, it’s just not affordable to do so. We’re really, we really want to create a good product for those. Smaller, um, smaller trips. Grocery is also very local, right? Like most people buy from a Grocery store that’s in their neighborhood. So, um, our form factor also is a really good vehicle for doing those sort of trips. And Grocery stores are a great example. It’s, you know, it’s the high volume hub and spoke. It’s a great example of where the size of the vehicle and the compactness of the vehicle is really important. You have a Grocery store in the middle of a city, you just can’t have shy volume of autonomous cars coming through. It. It’ll be chaos. Um. So we, we can kind of manage that chaos much better with this sort this sort of vehicle.

Grayson Brulte: It fits. But you know the other thing that. Your bots enable with Grocery stores is fresh food. I go to the Grocery store every day. I drive go there ’cause I wanna buy it fresh, but yet if I had a delivery bot, it makes it much easier and I have more time at home. So that, that’s an advantage there because too many individuals buy stuff full of pesticides and chemicals. Oh, it’s gonna last a week. Well, and your insight, it’s not very good. It’s sitting in there rotting. But if you can have fresh. It’s lower healthcare costs. You can think better. It’s just better for everybody involved. I’m a big fan of what you’re building. Zach, we wish you well. I can’t wait to see how you scale when you hit this 10,000 bot. We’re gonna have you back in the road to autonomy next year as you scale, as you prepare to scale, obviously as a founder and the CEO, you’re thinking about the future. What is the future of Coco Robotics? 

Zach Rash: Yeah, I think right now we’re. Really, there’s a, there’s a lot of work to do to capture kind of this initial product offering, which is a autonomous vehicle for moving goods in a city, right? That is a, that is a, an enormous market. So that’s gonna take a lot of our attention for a while. I think, uh, a stat that I say often that I think catches people off guard is, uh, DoorDash reports. I think, uh, around 16 billion a year is what the kind of career earnings are on their platform. So it’s like you got one, one company mostly in the US delivering mostly hot food at the current, current price point being that expensive and that is generating like 16 billion in courier payments. So that’s, it’s a pretty insanely large market, and I think if you bring that cost down substantially, that’s going to go way higher. I, guys maybe have a better analysis on this than me, but I would assume like. If you include the rest of the US, the companies in the US and Europe, you’re probably, dunno, 40 to $50 billion of career courier payments for, for moving food and Groceryeries. So. , So we wanna really drive the cost way down, get the cost per delivery to be really low. Um, so that’s more affordable for people . and uh, and that’s gonna take, that’s gonna take a lot of vehicles. It’s gonna take a lot of countries. Uh, so we’re really focused on that. But I think once you do that, there’s a lot of opportunity for, um, you know, if you have these automated vehicles in the cities, there’s a lot of ways to benefit the community, to benefit the cities, to transport all types of goods, retail goods. Pharmacy goods, um, and, uh, and offer all of those in an ultra low cost. So I think we’ll always, uh, we’ll be, we’ll be focused on that North star of making things much more accessible to people. Um, but there’s a lot, there’s a lot of work to do to do that.

Grayson Brulte: There’s a lot of work to do, but what we do know is that robot delivery is a growing market. The future is bright, the future is autonomous. The future is Coco Robotics. Zach, thank you so much for coming on The Road to Autonomy today.

Zach Rash: Yeah. Thanks for having me, Grayson.

Key The Road to Autonomy Episode Questions Answered